In Part 1, I shared what my research into 71 mining fatalities in Western Australia found: fatalities rarely have a single cause, the real causes start upstream, and many of those causes never make it into the data.
In this part, I want to zoom out even further, because some of the most important conditions shaping safety performance sit outside the organization altogether.
This blog covers how safety performance can shift with the economy, what subcontracting means for who carries the risk, what the ASTM E2920-26 SIF standard changes, why AI can scale existing data blind spots, and practical steps for building a serious injury and fatality (SIF) prevention program that looks at risk both internally and externally.
Worker, Internal and External Factors in Safety Risk
The best safety programs assess risk at three levels. They look at the worker. They look at internal factors, those a company can control: are risk assessments accurate and done in time, is equipment being serviced in a reasonable period? And they look at external factors, the ones an organization can’t directly control but that still shape safety performance. Is the economy in a boom or a recession? Are there climate events?
We often treat the internal factors as the complete job, and we’re comfortable reporting on them. External factors get treated as noise, or worse, explained away. ISO 31000 tells us to build both into our risk framework, yet most organizations are only using the internal part. Not being able to control something doesn’t mean we can’t detect and manage it.
How Economic Cycles Affect Workplace Injury Rates
I analyzed 577,778 injury records across Western Australia over 17 years. Injury rates moved with the economic cycle, with a big drop around the global financial crisis. For context, mining accounts for 43% of Western Australia’s GDP, with transient, mobile crews spread across a vast, remote state.
Then I kept seeing the same pattern challenged by industry. During economic growth periods, injury rates in mining and construction fell as employment grew. In manufacturing, they rose. Same economic conditions. Opposite outcomes.
My reading is that it comes down to how prepared each industry was to absorb new workers. Mining and construction both rely on rigorous, formal safety induction for new starters. That difference, not the injury rates themselves, is what the research is really about: how organizations respond to changing economic conditions, and how those responses shape safety. When we rely on injury metrics to predict fatalities, we miss the point.
Subcontracting and Safety: Who Carries the Risk?
I also found that industries with a high contractor workforce were more sensitive to economic change than larger organizations with the resources to be more resilient.
The pattern I found assumes the organization experiencing the economic pressure is the same one managing the safety risk. Subcontracting breaks that assumption. The organization making the decision is not always the one carrying the risk. And the one carrying the risk doesn’t always have the information to recognize when production pressure has changed the conditions they’re working under.
ASTM E2920-26: The Standard for SIF Classification
For years, the industry became skilled at counting how often incidents occur. We never had a consistent way to measure how close we came to a catastrophe. That gap is where fatalities are hidden.
The release of ASTM E2920-26 changes that. It’s the most significant update to SIF classification the industry has seen. It’s voluntary, not a regulation, but it’s the first time the industry has a shared foundation for real benchmarking. That consistency matters even more as organizations start applying AI to their safety data.
Infographic: Download Serious Injuries and Fatalities: SIFs and How to Prevent Them
AI in EHS: Why Data Quality Comes First
AI doesn’t fix your problems; it scales them. If the data is already blind to the systemic causes, you get something very confident, very fast, and still looking in the wrong place.
It’s not about better algorithms. It’s about widening what we let the system see: the conditions, not just the incident count. Get that right, and AI stops scaling your blind spots and starts helping you interrupt the risk before it reaches the worker.
Practical Steps to Build Your SIF Prevention Program
If we’re going to bring fatalities down, we need to look through both lenses.
Internally
- Identify the pathways driving your fatalities. Before designing any program, understand where your risks are. Look at the data you already have and ask which pathways keep appearing, which rarely appear, and what that tells you about how risk is being understood.
- Validate your controls. Confirm that the controls preventing serious harm are real and actually being used in the field. Tools like the energy wheel and high energy control assessments (HECA), developed by the U.S. electrical utility sector, help you see high-consequence risks earlier.
- Classify consistently. When you have thousands of events, the safety classification and learning (SCL) model, tested on around 14,000 incidents, lets you classify every event the same way, with the same risk language and severity in every region. That gives you a targeted list to prioritize and work through.
Externally
This is where the remaining risk sits, and where I think we can make the most difference.
- Make reporting safe and close the loop. Warnings only help if people feel safe raising them and see a visible response when they do. Look beyond near misses to “far misses” too: just because you haven’t had an incident yet doesn’t mean you won’t.
- Hold joint committees. Bring frontline workers together with managers and supervisors for a more complete picture of risk.
- Watch the conditions. You can’t control the economy, but you can watch your induction, training and contractor workforce as conditions change.
Final Thoughts
Strong safety systems have taken us a long way, but many were focused on preventing injuries, leaving organizations surprised when a fatality occurred. What happens beyond the organization, from shifts in the economy to who is doing the work and the tools we use to make sense of it, shapes safety too. The organizations that make real progress will be the ones that build those conditions into how they assess safety. Start with the data you already have, and ask what it isn’t showing you.
References
Bautista-Bernal, I., Quintana-García, C., & Marchante-Lara, M. (2024). Safety culture, safety performance and financial performance. A longitudinal study. Safety Science, 172, 106409. https://doi.org/10.1016/j.ssci.2023.106409
International Labour Organization. (2024). Ensuring safety and health at work in a changing climate. International Labour Organization.
International Organization for Standardization. (2018). ISO 31000 – Risk management. International Organization for Standardization. https://www.iso.org/obp/ui/#iso:std:iso:31000:ed-2:v1:en
Jenke, T., Oosthuizen, J., & Cattani, M. (2021). An investigation of the influence of economic cycles on safety performance in Western Australia. Safety Science, 138, 105230. https://doi.org/10.1016/j.ssci.2021.105230
